Draft
Conversation
Brings over the multihost inference work from fix/simpo-multihost-inference, excluding all SimPO/preference dataset changes. Core changes: - Inference engine overhaul: page-based KV cache, multi-round generation, DCN axis handling, sync barriers - JIT scheduler for paged attention with page allocation - Multihost sampling entrypoint (sample_lm_multihost.py) with per-round reset, host-level data parallelism, prompt sharding, WandB logging - Attention layer changes for paged KV cache and TPU RPA support - KV cache restructuring for physical page management across hosts - 180 sampler config YAMLs organized by milestone (m2-m11) - Tests for engine, scheduler, page table, KV cache, and multihost
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The main document to look at is lib/levanter/CODEX_REFACTOR_KV.md this is where I had codex write high level plans I would approve.
it would then write more details and logging to lib/levanter/CODEX_INFERNCE_MXX.md where MXX is milestone M1 M2 etc. Seems to work!
sample command:
python infra/launch.py
--tpu_name
--tpu_type v5p-32
--node_count 4
--zone us-central1-a
-- python -m levanter.main.sample_lm_multihost
--config_path lib/levanter/config/sampler/sample_llama8b_multihost_real_128prompts_2048_m10_hostdp_wandb_v5p32.yaml